精子多目标跟踪的抗遮挡算法的研究与实现  

Research and implementation of anti-occlusion algorithm for sperm multi-target tracking

作　　者：冯汉港 张艳青[1] 郑运平[1] 

机构地区：[1]华南理工大学计算机科学与工程学院,广东广州510006

出　　处：《计算机工程与设计》2014年第12期4380-4385,共6页Computer Engineering and Design

基　　金：国家自然科学基金项目(61300134)

摘　　要：为解决计算机辅助精液分析(CASA)系统中原版精子多目标跟踪算法在精子被白细胞或其它精子遮挡时,会产生检测失败和跟踪错误等问题,提出一种改进的抗遮挡算法。根据精子目标的面积和形态特征,判定遮挡的类型和发生期间,利用Kalman滤波对白细胞遮挡的精子进行虚拟定位跟踪,对精子互相遮挡的目标保持特征信息跟踪,在遮挡结束时重新匹配找回目标。实验结果表明,改进的算法能够成功跟踪到遮挡情况下的精子目标运动轨迹,提高了精子多目标跟踪算法的准确性和鲁棒性。To solve the problems such as detecting failures and tracking errors occurred in the original sperm multi-target tracking algorithm of computer aided sperm analysis（CASA）system when sperm sperms are occluded by white blood cells or other sperms,an improved anti-occlusion algorithm was presented.Based on the sperm target’s size and morphological characteristics,the type of occlusion and the occurring period were determined.Then sperms occluded by white cells were tracked with virtual positioning using the Kalman filter,and sperm targets occluded by each other were tracked with retentive information.Back targets were matched again at the end of occlusion.Experimental results show that the improved anti-occlusion algorithm can successfully track the sperm targets’ moving trajectories when occlusion occurring,increasing the accuracy and robustness of sperm multi-target tracking algorithm.

关 键 词：抗遮挡 目标识别 KALMAN滤波 多目标跟踪 模板匹配 计算机辅助精子分析 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]